Someone will no doubt do the maths and tell us which is which, you can do it with a spinning bike wheel and see for yourself.
Easy way to remember it:
Gyroscopic action causes the motion in response to a force to be "carried along" with the rotation by a quarter-turn.
So if, for instance, the blade is turning clockwise (as viewed from the front) and when going into furl is yawed clockwise (as viewed by from the top), the forward push at 3 oclock becomes a forward motion of the blade at 6 oclock, deflecting the blade away from the tower.
This is because the force on the blade causes an ACCELLERATION of its motion. In the above scenario the force toward the wind on a given blade starts at zero at noon, becomes maximum at 3, and goes back to zero at 6 - accellerating the blade toward the windward all the way. So when it gets to 6:00 it's moving maximally into the wind in response to the force. (Then the same thing happens in reverse on the other half turn.)
As Flux said:
It is likely that in operation, the yaw into furl will be more violent than the recovery from furl and it should make sense to have the yaw towards furl bring the blade tips clear of the tower.
This is because unfurling force is limited by the force from the tail, which is in turn limited by the furling system geometry (pivot angles, tail length, weight, and vane area) and is in opposed by the force on the offset turbine, which nearly balances it. Too much force from the wind and the tail folds up and the mill furls again. Furling force is limited only by the strength of the gust (which is arbitrarily large), and with the tail having bailed out almost completely in a strong gust the force from the offset turbine is nearly unresisted. Also: Furling happens when the turbine speed is near maximum and accellerating, maximizing gyroscopic effect, while unfurling occurs when the prop is slowing down from a furling-limited speed (which may be lower).